Cultures of Fetishism (35 page)

No matter how smart Kismet becomes, no matter how adept it becomes at responding to its “mother’s” queries and instructions, it is after all a mechanical thing of metal parts, computer-regulated motors, and wires.

Kismet’s most prominent features are his two large eyes with two foveal- front vision cameras set up behind their pupils. Two wide-angle, peripheral vision cameras are located behind the space where a nose would be. It has microphones in its ears. There is a gyroscope in the middle of its head. Aside from this basic sensory equipment, Kismet has several motors that control its eyelids and eyebrows, which can go up and down and sideways; four motors that guide the thick gray metal plates to move synergistically with the shiny red lips that can spread apart and open and close and bend to one side or another; and floppy doglike ears that can go up and down and forward and backward and sideways. It has built-in actuators that allow it to move its neck around three different axes and its eyes to scan left to right and up and down.
14

Breazeal has created an ingenious robot, which, because it is designed to behave as if it has human thoughts and feelings, is technically an android even though it has no body or arms or legs and has a face that is a grotesque imi- tation of a human infant’s face. Much of what Breazeal predicts about Kismet’s future and the quality of its relationship to her is pure fantasy; a Pygmalion-Galatea fantasy translated into mommy-baby language.

As Breazeal explains in her book
Designing Sociable Robots
, Kismet was designed to have three basic “drives.” Yes, drives—those innate energies and inner vitalities that have always eluded the grasp of human intelligence. Are such enigmatic “living” faculties possible in machines made of metal parts and energized by motors and computers? After I became more familiar with the energy sources activating Kismet’s sensory equipment and motor responses, I could appreciate that an android, even one with a barely human face, could be fabricated to have artificially created drives, and that further- more these so-called drives would also make the robot appear to be alive— expecially if the human who interacts with it wants to believe it is alive. If Kismet is to come to life, the human who chooses to interact with it has to bring a fantasy of robotic aliveness to the situation.

It seems that humans have a hubris drive that inspires them with the fan- tasy that they are able to breathe life into inanimate objects. Humans also have an anthropomorphic drive that motivates their fantasies that certain objects in the world around them are endowed with human characteristics. The most efficient social robot has features built into it that encourage humans to fulfill their fantasies by personifying it and believing that it is responding to them like a human being. As Breazeal readily admits, human beings who confront Kismet for the first time have to be trained to appreci- ate the meanings of Kismet’s facial expressions and head movements. Only then are they able to elicit humanlike behaviors in Kismet.

Therefore the first of Kismet’s drives,
15
an essential for designing sociable robots, is a
social
drive, a drive that motivates the robot to want to be in the

presence of people who will interact with it. When Kismet is deprived of human interaction, it falls into a state that Breazeal has labeled “loneliness.” The state of loneliness then motivates Kismet to behave in ways that help it to establish face-to-face contact with humans. On the other hand, if a human moves too close to Kismet or positions herself too close to its eyes, Kismet will feel overwhelmed and activate movements that allow it to avoid face-to- face contact.

The second drive, obviously tied to the first, is a stimulation drive.
16
This drive can be gratified if a human engages the robot’s attention with a color- ful toy. If Kismet is not stimulated over a period of time, it gets “bored,” and displays this boredom by moving its eyes and mouth and ears into a posture of “please stimulate me.” If Kismet is given too much stimulation, that is, more stimulation than its perceptual processes can tolerate, it will bring its eyelids down over its eyes, or turn its head away from the source of stimula- tion. The human is thereby encouraged to challenge the robot to engage in new activities; but not too much, and not too many all at one time. Humans learn to gauge and regulate their responses in ways that activate the social qualities of Kismet. As Breazeal puts it, “an ongoing dance between robot and human”
17
maintains the robot’s drives.

Coming to the rescue from all this drive stimulation is Kismet’s third drive, fatigue.
18
Eventually, as time passes and the fatigue drive approaches the exhausted level, Kismet “goes to sleep,” and all drives return to home- ostasis. The fatigue drive allows Kismet to shut out the external environment and to terminate any further responses.

When Kismet is awake, it is open to receiving repeated stimulations from the environment and able to respond to them with “emotions” such as anger and frustration, or joy and fear, which are accompanied by changes in facial expression that motivate the caregiver to modify her behavior.
19
For example, when Kismet displays a sorrowful facial expression (ears down, eye- lids half over the eyes, eyebrows and lips turned slightly downward), it evokes sympathy and attention from the caregiver. Kismet shows its boredom by moving its head back and forth, lifting its ears, turning them outward and opening its eyes, in order to motivate humans to supply it with external stimulation.

Unlike the android child, David, in the science fiction film
Artificial Intelligence
, or the vast majority of androids that are being constructed these days, baby Kismet has not been programmed in advance by a set of rules that require it to respond in specific ways to specific environmental stimuli. Like most of the other robots in Brooks’ lab, many of which have been con- structed in the shape of insects, Kismet is designed so it can learn how to act and react from its interactions with the environment and thereby acquire or learn new responses. Rather than having an intelligence that is pre-set and automatic, its intelligence emerges as it interacts. Technically, it is called “emergent Artificial Intelligence.”

On the surface, at least, this is similar to the way human intelligence devel- ops. We come into the world with certain response capabilities. But soon

after birth we begin to learn through our interactions with our environment. Intelligence can only come into existence in an environment that encourages interaction with it. However, unlike the human baby who also learns by inter- acting with his environment, Kismet does not have a central controlling apparatus—a primed-to-respond cognitive system or brainlike, nervous system mechanisms to guide and regulate its sensory-motor behaviors. Instead of being constructed with this kind of top-down arrangement, Kismet was designed to have many independent but closely connected visual and auditory systems that would add up to enabling Kismet to seem as if it could think. In artificial intelligence (AI) language, this is sometimes referred to as a bottom-up form of activation and control.

Brooks believes that this sort of bottom-up, embodied, situated-in-the- world type of intelligence will eventually bring robotic intelligence up to par with human intelligence.
20
Nevertheless, even he can appreciate the consid- erable limitations of Kismet’s conversational style. “What Kismet cannot do is actually understand what is said to it. Nor can it say anything meaningful.”
21

For Breazeal, the ultimate, grand challenge of robotic life is the construc- tion of “
anima machina
.”
22
This is a fantasy, an extension of Breazeal’s Pygmalion-Galatea fantasy of bringing inanimate substances to life. Even the hard-nosed Brooks, who appreciates the limitations of Kismet and under- stands full well the provocative nature of his statement about machines becoming human and carbon-based human DNA “machines” that can one day cross the border into becoming silicon machines, sometimes succumbs to his own more elaborate Pygmalion fantasy of “living, breathing robots.”
23

Merely to begin to meet this challenge of imbuing robots with life energies, robot scientists would have to be able to construct an android that can manage its own daily physical existence without the assistance of humans. The android would have to have a synthetic nervous system that could “breathe the life” into the machine that is it. The android would have to have the ability to carry its own energy source wherever it goes and the ability to replenish this energy over time. Basically it would have to have access to bat- teries or electrical plugs and understand how to connect itself up with these sources of energy. To be entitled to the name “android,” a robot would have to be more than a head, or a head and arms. It would have to be bipedal and capable of walking about in the world.

During the years that Brooks and Breazeal were working diligently to breathe real life into their mechanical robotic creations, other robot scientists had already taken a more practical approach to robotic intelligence. If they had any fantasies about breathing life into the machines they constructed, they did not let it interfere with their engineering ingenuities.

Within two decades, Japan, which had been losing out to the United States in the computer business, became Land of the Robots. After twelve years of experimentation by thirty dedicated research engineers, and many millions of dollars, and three preliminary prototypes, the Honda Motor company suc- ceeded in building the first bipedal robot.
24
In 1993, they produced P1, the

first prototype. Three years later, at a cost of $105 million more, they pro- duced P2, which could walk like a human and “even climb up and down stairs.”
25
P3, which arrived on the scene a year later, could also climb stairs and, in addition, could walk at the same speed as a human, kneel down and then stand up straight. When jostled it could regain its balance.
26

And then in 2001, the bipedal android the world had been waiting for was introduced. It was given the name Asimo. Many people assumed Asimo had been named for Issac Asimov. However, Asimo, a Japanese android, was both the acronymn for Advanced Step in Innovative Mobility, and also the combi- nation of
asi
, the Japanese word for feet, plus
mo
, the first two letters of the word for move. Asimo has feet that move.
27

Unlike its predecessors, Asimo had been designed with interactive capaci- ties. It could understand facial expressions and gestures and obey spoken orders. If somebody said, “Follow me,” it would do so. If someone asked it to look for something, it would do so and also find it. It knew the names that went with different faces and could even go online to retrieve news and other information like the weather forecast.
28

It will be quite some time before Asimo earns back what it cost Honda to invent and construct it. However, by 2004, there were thirty Asimos in the real world handling real jobs of various sorts. For the nifty price of $175,500, other industrial companies could lease an Asimo for a year. IBM, for example, rented an Asimo to act as an obedient and very prestigious high-tech receptionist.
29

Sony, stimulated by Honda’s successes, didn’t lag far behind its mightiest competitor. Sony focused on money-making robots that could serve as com- panions in the home. One of their star robotic engineers, Toshida Doi, had once worked with Brooks at MIT. Said Doi of Brooks’ emergent AI philoso- phy, “Brooks was too minimalist.”
30
As a result there were limits on the pos- sible practical applications of his emergent AI research methodology. Soon after breaking out of the Brooksian mold, Doi struck out on his own to create a recreational household pet—his famous robotic dog, Aibo
31
which means “friend” in Japanese. Doi, whose robotic ideals are different from those of Brooks, believes in programming robots from the top down and making sure their responses are well-controlled, so that they can appeal to the humans who will buy them and make them a part of their household. Although he designs his robots to behave in a “lifelike” manner, he is not interested in breathing actual life into his machines. He speaks of his creations as works of art. What his fantasies are about the robots he creates, he does not say. But I imagine they have something to do with “The Dreams that Money Can Buy.”

By 1999 Sony was marketing Doi’s robotic dog all over the world and rak- ing in big bucks. In 2003, the Aibo upgrade, black pearl Aibo, appeared on the market. This entertaining robot dog could turn on CD’s, collect e-mails, and act as a watchdog or an assistant who reminded its master of his daily work schedule. Sony sold hundreds of thousands within weeks.
32
When Doi, who by then had become president of Sony’s Entertainment Robot Company, expressed a wish to build an android that could be an entertaining home companion, he received Sony’s unqualified blessing and all the funding

he requested. Near the end of 2003, Sony introduced Doi’s tiny two-foot-tall Qrio, who could do all the things the four and half-foot Asimo did, such as interact with humans and hold conversations. But, Qrio could also move its arms and legs to the beat of music and throw a tiny football.
33
Despite its tiny dimensions, Qrio was more than an entertaining toy. Technically it could be considered an android. However, like Asimo, Kismet, and the vast majority of other As-If human robots, Qrio does not have a human face and does not look like any of the human-looking androids that we are used to seeing in science-fiction films.

NEC invented Pa Pe Ro, Partner Personal Robot, a home companion that was designed to seek out interaction with humans. It has a tiny, brightly colored body, an adorable chubby face with huge eyes that can light up. It dances, tells time, can switch on the TV, and go online to pick up and forward messages. If Pa Pe Ro is treated with affection it will be gentle and helpful. A harsh tone of voice will make it grumpy and lazy. It recognizes 50,000 Japanese words and can speak over 3,000 phrases.
34
But it does not look like a human being.